How the Mammoth data quality score is calculated
The Data Quality Report gives a View an overall score out of 100 from six DAMA (Data Management Association) dimensions, each scored 0–100%. You see it in the Overview of the report: click View Data Quality in the footer below the data grid, then View Details. How to run and read the report is on Data quality report.
The six dimensions and their weights
Dimensions count columns with a problem, not rows, so one bad column lowers a dimension by its share of the columns checked. The overall score is the weighted average, rounded down to a whole number.
| Dimension | Weight | What it measures | How it's scored |
|---|---|---|---|
| Validity | 25% | Values unexpected for what the column holds | Share of columns without unexpected negatives (prices, quantities), unexpected zeros, or future dates. Text formats such as email or phone patterns are not checked |
| Uniqueness | 25% | Duplicates in columns that should be unique | Share of identifier columns (IDs and keys) without duplicates. A column is flagged when it is less than 95% unique. No identifier columns scores 0% (a dash) |
| Completeness | 20% | Nulls, empty fields, empty columns | Share of columns with no missing values |
| Consistency | 15% | Case variants in text columns, such as "Active", "ACTIVE" and "active" | Share of text columns without case variants. No text columns scores 100%. Relationships between columns are not checked |
| Accuracy | 10% | Numeric outliers, found with the interquartile range (IQR) method | Share of number columns without outliers. Never below 60%, because outliers need domain knowledge to judge. No number columns scores 100% |
| Timeliness | 5% | Age of the newest date against the refresh frequency the AI expects (for example, daily for transactions) | 100% within the expected age, 70% up to twice it, 50% up to four times, 25% beyond. With no expected cadence: 100% up to 30 days old, 70% up to 90 days, 50% beyond. No date columns scores 100% |
The report states the measured age and, when there is one, the expected cadence, for example "Data is 1678 days old (very stale vs expected 1 day)". Common causes of stale data are ETL jobs failing silently and sources that stopped sending.
Why a dimension shows a dash
A dimension that scores 0% shows a dash ("-") instead of a number. For Uniqueness, this happens when every identifier column has duplicates, or when the View has no identifier columns. The note under the dash says which.
Because dimensions count columns, a View with few columns swings sharply, and the weights may not match your priorities. Read the dimension tiles and the issue cards, not only the total.
Score levels: Excellent, Good, Fair and Poor
| Score | Level | What it means | Typical next step |
|---|---|---|---|
| 95–100% | Excellent | Production-ready; minor formatting issues at most | Publish dashboards, share, automate reporting; record the score for audit or compliance; set up an automation to refresh the data on a schedule |
| 80–94% | Good | Acceptable with awareness: some missing values, formatting or minor duplicates | Note the limits; consider minor cleanup |
| 60–79% | Fair | Several dimensions affected, some critical issues | Fix critical issues before production; development use only |
| Below 60% | Poor | Major problems; results may be unreliable. Quality assessment only, not suitable for analysis | Fix critical issues first; don't proceed to analysis until quality improves; consider fixing at the source |
Related
- Data quality report: generate the report and read its sections.
- Fix common data quality problems: the pipeline steps that raise each dimension.
- Schedules and triggers: refresh data on a schedule so Timeliness stays high.
Something wrong? support@mammoth.io